【作者】 杨倩；

【导师】 孔德信；

【作者基本信息】 华中农业大学 ， 生物信息学， 2018， 硕士

【摘要】 肿瘤是典型的复杂性疾病,采用多靶标的设计策略才能解决化疗药物的耐药性问题,降低药物的使用浓度,减少毒副作用。多靶标药物设计可分为两个步骤:靶标组合的确定和多靶标药物的筛选与设计,前者是多靶标药物设计的关键。本论文基于NCI数据库中化合物对常见肿瘤细胞系的生长抑制活性数据以及Binding DB数据库中不同靶标的大量活性分子数据,采用两层相似性融合算法(Two-Layer Similarity Ensemble Approach,TL-SEA)建立了NCI抗肿瘤化合物与Binding DB靶标的相互作用网络,最终实现了抗肿瘤靶标组合的预测。首先,从NCI下载具有抗肿瘤活性的小分子数据,从BindingDB下载不同靶标的小分子活性数据,经保留分子最大片段,加氢,理化性质筛选等的预处理后,计算了NCI抗肿瘤活性分子和Binding DB靶标活性分子的相似性矩阵。然后,采用TL-SEA算法,计算NCI小分子与靶标之间的关联度打分(P值),建立了抗肿瘤活性小分子和Binding DB靶标蛋白的相互作用网络。再后,根据NCI活性分子对任意两个Binding DB靶标的关联度打分(P值)向量,定义任意靶标对之间的组合可信度打分。循环计算所有的靶标组合的可信度打分Q,获得潜在的抗肿瘤靶标对组合。进一步,计算了Binding DB数据库中靶标蛋白之间的关联打分,将上述获得的抗肿瘤靶标对组合中的同类蛋白或配基结构相似的蛋白剔除。最后,利用上述算法计算可购买化合物库Specs中的小分子与保留的抗肿瘤靶标对之间的关联打分,选取具有高打分的小分子,以组合或者单独的形式进行肿瘤细胞生长抑制实验,验证该靶标组合的有效性。结果表明有6个化合物,对K-562细胞系具有较高的生长抑制活性。本论文基于公开的生物活性大数据以及可靠的药物发现算法,建立了肿瘤靶标组合和药物筛选流程。论文研究有助于抗肿瘤多靶标药物的发现,而且还对糖尿病、老年痴呆等复杂性疾病的多靶标药物发现提供理论方法参考。更多还原

【Abstract】 Tumor is a typical complex disease.Using multi-target design strategy can help to solve the drug resistance problem caused by traditional chemotherapy,reduce the drug concentration and undesirable side effects.The process of multi-target drug design involves two steps,identification of the target combinations and the multi-target drug design,the former is the key in multi-target drug design.In this thesis,based on the anti-tumor data in NCI and activity data on the various targets in Binding DB,we built the interaction network of NCI anti-tumor compounds and Binding DB targets with the Two-Layer Similarity Ensemble Approach(TL-SEA)algorithm and successfully predicted the anti-tumor target combinations.First,the structure and activity data of anti-tumor small molecules in NCI and of the ligands in Binding DB were downloaded.Preprocessing,including keep the largest fragment,add hydrogens,filtering with physicochemical property and activity criteria,the similarity matrix of NCI anti-tumor active molecules and active ligands against the Binding DB targets were calculated.Then,the TL-SEA algorithm was utilized to calculate the association value(P)between the NCI small molecules and the target.The interaction network of the NCI small anti-tumor molecules with the Binding DB targets were constructed.Then,based on the P value vectors between the all NCI active molecules to a pair of Binding DB targets,we defined the combination confidence score (Q) of the target pair as.The combination confidence scores(Q)of all possible BindingDB target combinations were calculated.Then,homology target pairs or similar proteins were excluded by calculating the P value between proteins in Binding DB with TL-SEA algorithm.Finally,available compounds supplied by Specs company were screened against all the valid target pairs.The tumor cell growth inhibition experiment was performed.The results indicated that six of the candidate compounds had inhibitory activities at micromole concentration on K-562 cell line.Corresponding target combination related to K-562 cell line were obtained.The research reported in this thesis provided a novel algorithm for target combination prediction based on the big data of bioactivity data and phenotype activities.The results will not only promote the discovery of anti-tumor multi-target drugs directly,but also provide a theoretical approach for the multi-target drug discovery of other complex disease,such as diabetes and Alzheimer’s disease.更多还原